I wrote the inference step of XGBoost in Go. It will make predictions after loading in an XGBoost JSON model. Writing the training portion of XGBoost would be much harder.